Merge branch 'bpf-sockmap-fix-data-loss-and-panic-issues'

Jiayuan Chen says: ==================== bpf, sockmap: Fix data loss and panic issues I was writing a benchmark based on sockmap + TCP and discovered several issues: 1. When EAGAIN occurs, the direction of skb is incorrect, causing data loss when retry. 2. When sending partial data, the offset is not recorded, leading to duplicate data being sent when retry. 3. An unexpected BUG_ON() judgment in skb_linearize is triggered. 4. The memory of psock->ingress_skb is not limited by the socket buffer and memcg. Issues 1, 2, and 3 are described in each patch's commit message. Regarding issue 4, this patchset does not cover it as it is difficult to handle in practice, and I am still working on it. Here is a brief description of the issue: When using sockmap to skb/stream redirect, if the receiving end does not perform read operations, all data will be buffered in ingress_skb. For example: ''' // set memory limit to 50G cgcreate -g memory:myGroup cgset -r memory.max="5000M" myGroup // start benchmark and disable consumer from reading cgexec -g "memory:myGroup" ./bench sockmap -c 2 -p 1 -a --rx-verdict-ingress --delay-consumer=-1 -d 100 Iter 0 ( 29.179us): Send Speed 2668.548 MB/s (20360.406 calls/s), ... Rcv Speed 0.000 MB/s ( 0.000 calls/s) Iter 1 ( -7.237us): Send Speed 2694.467 MB/s (20557.149 calls/s), ... Rcv Speed 0.000 MB/s ( 0.000 calls/s) Iter 2 ( -1.918us): Send Speed 2693.404 MB/s (20548.039 calls/s), ... Rcv Speed 0.000 MB/s ( 0.000 calls/s) Iter 3 ( -0.684us): Send Speed 2693.138 MB/s (20548.014 calls/s), ... Rcv Speed 0.000 MB/s ( 0.000 calls/s) Iter 4 ( 7.879us): Send Speed 2698.620 MB/s (20588.838 calls/s), ... Rcv Speed 0.000 MB/s ( 0.000 calls/s) Iter 5 ( -3.224us): Send Speed 2696.553 MB/s (20573.066 calls/s), ... Rcv Speed 0.000 MB/s ( 0.000 calls/s) Iter 6 ( -5.409us): Send Speed 2699.705 MB/s (20597.111 calls/s), ... Rcv Speed 0.000 MB/s ( 0.000 calls/s) Iter 7 ( -0.439us): Send Speed 2699.691 MB/s (20597.009 calls/s), ... Rcv Speed 0.000 MB/s ( 0.000 calls/s) ... // memory usage are not limited cat /proc/slabinfo | grep skb skbuff_small_head 11824024 11824024 704 46 8 : tunables 0 0 0 : slabdata 257044 257044 0 skbuff_fclone_cache 11822080 11822080 512 32 4 : tunables 0 0 0 : slabdata 369440 369440 0 ''' Thus, a simple socket in a large file upload/download model can eat the entire OS memory. We must charge the skb memory to psock->sk, and if we do not want losing skb, we need to feedback the error info to read_sock/read_skb when the enqueue operation of psock->ingress_skb fails. --- My another patch related to stability also requires maintainers to spare some time from their busy schedules for review. https://lore.kernel.org/bpf/20250317092257.68760-1-jiayuan.chen@linux.dev/T/#t ==================== Link: https://patch.msgid.link/20250407142234.47591-1-jiayuan.chen@linux.dev Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2026-05-06 07:34:03 -04:00 · 2025-04-09 19:59:00 -07:00
parent 727d00a51f 7b2fa44de5
commit 690d43d3b3
5 changed files with 697 additions and 21 deletions
--- a/net/core/skmsg.c
+++ b/net/core/skmsg.c
@@ -530,16 +530,22 @@ static int sk_psock_skb_ingress_enqueue(struct sk_buff *skb,
 					u32 off, u32 len,
 					struct sk_psock *psock,
 					struct sock *sk,
-					struct sk_msg *msg)
+					struct sk_msg *msg,
+					bool take_ref)
 {
 	int num_sge, copied;

+	/* skb_to_sgvec will fail when the total number of fragments in
+	 * frag_list and frags exceeds MAX_MSG_FRAGS. For example, the
+	 * caller may aggregate multiple skbs.
+	 */
 	num_sge = skb_to_sgvec(skb, msg->sg.data, off, len);
 	if (num_sge < 0) {
 		/* skb linearize may fail with ENOMEM, but lets simply try again
 		 * later if this happens. Under memory pressure we don't want to
 		 * drop the skb. We need to linearize the skb so that the mapping
 		 * in skb_to_sgvec can not error.
+		 * Note that skb_linearize requires the skb not to be shared.
 		 */
 		if (skb_linearize(skb))
 			return -EAGAIN;
@@ -556,7 +562,7 @@ static int sk_psock_skb_ingress_enqueue(struct sk_buff *skb,
 	msg->sg.start = 0;
 	msg->sg.size = copied;
 	msg->sg.end = num_sge;
-	msg->skb = skb;
+	msg->skb = take_ref ? skb_get(skb) : skb;

 	sk_psock_queue_msg(psock, msg);
 	sk_psock_data_ready(sk, psock);
@@ -564,7 +570,7 @@ static int sk_psock_skb_ingress_enqueue(struct sk_buff *skb,
 }

 static int sk_psock_skb_ingress_self(struct sk_psock *psock, struct sk_buff *skb,
-				     u32 off, u32 len);
+				     u32 off, u32 len, bool take_ref);

 static int sk_psock_skb_ingress(struct sk_psock *psock, struct sk_buff *skb,
 				u32 off, u32 len)
@@ -578,7 +584,7 @@ static int sk_psock_skb_ingress(struct sk_psock *psock, struct sk_buff *skb,
 	 * correctly.
 	 */
 	if (unlikely(skb->sk == sk))
-		return sk_psock_skb_ingress_self(psock, skb, off, len);
+		return sk_psock_skb_ingress_self(psock, skb, off, len, true);
 	msg = sk_psock_create_ingress_msg(sk, skb);
 	if (!msg)
 		return -EAGAIN;
@@ -590,7 +596,7 @@ static int sk_psock_skb_ingress(struct sk_psock *psock, struct sk_buff *skb,
 	 * into user buffers.
 	 */
 	skb_set_owner_r(skb, sk);
-	err = sk_psock_skb_ingress_enqueue(skb, off, len, psock, sk, msg);
+	err = sk_psock_skb_ingress_enqueue(skb, off, len, psock, sk, msg, true);
 	if (err < 0)
 		kfree(msg);
 	return err;
@@ -601,7 +607,7 @@ static int sk_psock_skb_ingress(struct sk_psock *psock, struct sk_buff *skb,
 * because the skb is already accounted for here.
 */
 static int sk_psock_skb_ingress_self(struct sk_psock *psock, struct sk_buff *skb,
-				     u32 off, u32 len)
+				     u32 off, u32 len, bool take_ref)
 {
 	struct sk_msg *msg = alloc_sk_msg(GFP_ATOMIC);
 	struct sock *sk = psock->sk;
@@ -610,7 +616,7 @@ static int sk_psock_skb_ingress_self(struct sk_psock *psock, struct sk_buff *skb
 	if (unlikely(!msg))
 		return -EAGAIN;
 	skb_set_owner_r(skb, sk);
-	err = sk_psock_skb_ingress_enqueue(skb, off, len, psock, sk, msg);
+	err = sk_psock_skb_ingress_enqueue(skb, off, len, psock, sk, msg, take_ref);
 	if (err < 0)
 		kfree(msg);
 	return err;
@@ -619,18 +625,13 @@ static int sk_psock_skb_ingress_self(struct sk_psock *psock, struct sk_buff *skb
 static int sk_psock_handle_skb(struct sk_psock *psock, struct sk_buff *skb,
 			       u32 off, u32 len, bool ingress)
 {
-	int err = 0;
-
 	if (!ingress) {
 		if (!sock_writeable(psock->sk))
 			return -EAGAIN;
 		return skb_send_sock(psock->sk, skb, off, len);
 	}
-	skb_get(skb);
-	err = sk_psock_skb_ingress(psock, skb, off, len);
-	if (err < 0)
-		kfree_skb(skb);
-	return err;
+
+	return sk_psock_skb_ingress(psock, skb, off, len);
 }

 static void sk_psock_skb_state(struct sk_psock *psock,
@@ -656,11 +657,6 @@ static void sk_psock_backlog(struct work_struct *work)
 	int ret;

 	mutex_lock(&psock->work_mutex);
-	if (unlikely(state->len)) {
-		len = state->len;
-		off = state->off;
-	}
-
 	while ((skb = skb_peek(&psock->ingress_skb))) {
 		len = skb->len;
 		off = 0;
@@ -670,6 +666,13 @@ static void sk_psock_backlog(struct work_struct *work)
 			off = stm->offset;
 			len = stm->full_len;
 		}
+
+		/* Resume processing from previous partial state */
+		if (unlikely(state->len)) {
+			len = state->len;
+			off = state->off;
+		}
+
 		ingress = skb_bpf_ingress(skb);
 		skb_bpf_redirect_clear(skb);
 		do {
@@ -680,7 +683,8 @@ static void sk_psock_backlog(struct work_struct *work)
 			if (ret <= 0) {
 				if (ret == -EAGAIN) {
 					sk_psock_skb_state(psock, state, len, off);
-
+					/* Restore redir info we cleared before */
+					skb_bpf_set_redir(skb, psock->sk, ingress);
 					/* Delay slightly to prioritize any
 					 * other work that might be here.
 					 */
@@ -697,6 +701,8 @@ static void sk_psock_backlog(struct work_struct *work)
 			len -= ret;
 		} while (len);

+		/* The entire skb sent, clear state */
+		sk_psock_skb_state(psock, state, 0, 0);
 		skb = skb_dequeue(&psock->ingress_skb);
 		kfree_skb(skb);
 	}
@@ -1014,7 +1020,7 @@ static int sk_psock_verdict_apply(struct sk_psock *psock, struct sk_buff *skb,
 				off = stm->offset;
 				len = stm->full_len;
 			}
-			err = sk_psock_skb_ingress_self(psock, skb, off, len);
+			err = sk_psock_skb_ingress_self(psock, skb, off, len, false);
 		}
 		if (err < 0) {
 			spin_lock_bh(&psock->ingress_lock);
--- a/tools/testing/selftests/bpf/Makefile
+++ b/tools/testing/selftests/bpf/Makefile
@@ -811,6 +811,7 @@ $(OUTPUT)/bench_local_storage_create.o: $(OUTPUT)/bench_local_storage_create.ske
 $(OUTPUT)/bench_bpf_hashmap_lookup.o: $(OUTPUT)/bpf_hashmap_lookup.skel.h
 $(OUTPUT)/bench_htab_mem.o: $(OUTPUT)/htab_mem_bench.skel.h
 $(OUTPUT)/bench_bpf_crypto.o: $(OUTPUT)/crypto_bench.skel.h
+$(OUTPUT)/bench_sockmap.o: $(OUTPUT)/bench_sockmap_prog.skel.h
 $(OUTPUT)/bench.o: bench.h testing_helpers.h $(BPFOBJ)
 $(OUTPUT)/bench: LDLIBS += -lm
 $(OUTPUT)/bench: $(OUTPUT)/bench.o \
@@ -831,6 +832,7 @@ $(OUTPUT)/bench: $(OUTPUT)/bench.o \
 		 $(OUTPUT)/bench_local_storage_create.o \
 		 $(OUTPUT)/bench_htab_mem.o \
 		 $(OUTPUT)/bench_bpf_crypto.o \
+		 $(OUTPUT)/bench_sockmap.o \
 		 #
 	$(call msg,BINARY,,$@)
 	$(Q)$(CC) $(CFLAGS) $(LDFLAGS) $(filter %.a %.o,$^) $(LDLIBS) -o $@
--- a/tools/testing/selftests/bpf/bench.c
+++ b/tools/testing/selftests/bpf/bench.c
@@ -283,6 +283,7 @@ extern struct argp bench_local_storage_create_argp;
 extern struct argp bench_htab_mem_argp;
 extern struct argp bench_trigger_batch_argp;
 extern struct argp bench_crypto_argp;
+extern struct argp bench_sockmap_argp;

 static const struct argp_child bench_parsers[] = {
 	{ &bench_ringbufs_argp, 0, "Ring buffers benchmark", 0 },
@@ -297,6 +298,7 @@ static const struct argp_child bench_parsers[] = {
 	{ &bench_htab_mem_argp, 0, "hash map memory benchmark", 0 },
 	{ &bench_trigger_batch_argp, 0, "BPF triggering benchmark", 0 },
 	{ &bench_crypto_argp, 0, "bpf crypto benchmark", 0 },
+	{ &bench_sockmap_argp, 0, "bpf sockmap benchmark", 0 },
 	{},
 };

@@ -549,6 +551,7 @@ extern const struct bench bench_local_storage_create;
 extern const struct bench bench_htab_mem;
 extern const struct bench bench_crypto_encrypt;
 extern const struct bench bench_crypto_decrypt;
+extern const struct bench bench_sockmap;

 static const struct bench *benchs[] = {
 	&bench_count_global,
@@ -609,6 +612,7 @@ static const struct bench *benchs[] = {
 	&bench_htab_mem,
 	&bench_crypto_encrypt,
 	&bench_crypto_decrypt,
+	&bench_sockmap,
 };

 static void find_benchmark(void)
--- a/tools/testing/selftests/bpf/benchs/bench_sockmap.c
+++ b/tools/testing/selftests/bpf/benchs/bench_sockmap.c
@@ -0,0 +1,599 @@
+// SPDX-License-Identifier: GPL-2.0
+
+#include <error.h>
+#include <sys/types.h>
+#include <sys/socket.h>
+#include <sys/types.h>
+#include <netinet/in.h>
+#include <sys/sendfile.h>
+#include <arpa/inet.h>
+#include <fcntl.h>
+#include <argp.h>
+#include "bench.h"
+#include "bench_sockmap_prog.skel.h"
+
+#define FILE_SIZE (128 * 1024)
+#define DATA_REPEAT_SIZE 10
+
+static const char snd_data[DATA_REPEAT_SIZE] = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9};
+
+/* c1 <-> [p1, p2] <-> c2
+ * RX bench(BPF_SK_SKB_STREAM_VERDICT):
+ *	ARG_FW_RX_PASS:
+ *		send(p2) -> recv(c2) -> bpf skb passthrough -> recv(c2)
+ *	ARG_FW_RX_VERDICT_EGRESS:
+ *		send(c1) -> verdict skb to tx queuec of p2 -> recv(c2)
+ *	ARG_FW_RX_VERDICT_INGRESS:
+ *		send(c1) -> verdict skb to rx queuec of c2 -> recv(c2)
+ *
+ * TX bench(BPF_SK_MSG_VERDIC):
+ *	ARG_FW_TX_PASS:
+ *		send(p2) -> bpf msg passthrough -> send(p2) -> recv(c2)
+ *	ARG_FW_TX_VERDICT_INGRESS:
+ *		send(p2) -> verdict msg to rx queue of c2 -> recv(c2)
+ *	ARG_FW_TX_VERDICT_EGRESS:
+ *		send(p1) -> verdict msg to tx queue of p2 -> recv(c2)
+ */
+enum SOCKMAP_ARG_FLAG {
+	ARG_FW_RX_NORMAL = 11000,
+	ARG_FW_RX_PASS,
+	ARG_FW_RX_VERDICT_EGRESS,
+	ARG_FW_RX_VERDICT_INGRESS,
+	ARG_FW_TX_NORMAL,
+	ARG_FW_TX_PASS,
+	ARG_FW_TX_VERDICT_INGRESS,
+	ARG_FW_TX_VERDICT_EGRESS,
+	ARG_CTL_RX_STRP,
+	ARG_CONSUMER_DELAY_TIME,
+	ARG_PRODUCER_DURATION,
+};
+
+#define TXMODE_NORMAL()				\
+	((ctx.mode) == ARG_FW_TX_NORMAL)
+
+#define TXMODE_BPF_INGRESS()			\
+	((ctx.mode) == ARG_FW_TX_VERDICT_INGRESS)
+
+#define TXMODE_BPF_EGRESS()			\
+	((ctx.mode) == ARG_FW_TX_VERDICT_EGRESS)
+
+#define TXMODE_BPF_PASS()			\
+	((ctx.mode) == ARG_FW_TX_PASS)
+
+#define TXMODE_BPF() (				\
+	TXMODE_BPF_PASS() ||			\
+	TXMODE_BPF_INGRESS() ||			\
+	TXMODE_BPF_EGRESS())
+
+#define TXMODE() (				\
+	TXMODE_NORMAL() ||			\
+	TXMODE_BPF())
+
+#define RXMODE_NORMAL()				\
+	((ctx.mode) == ARG_FW_RX_NORMAL)
+
+#define RXMODE_BPF_PASS()			\
+	((ctx.mode) == ARG_FW_RX_PASS)
+
+#define RXMODE_BPF_VERDICT_EGRESS()		\
+	((ctx.mode) == ARG_FW_RX_VERDICT_EGRESS)
+
+#define RXMODE_BPF_VERDICT_INGRESS()		\
+	((ctx.mode) == ARG_FW_RX_VERDICT_INGRESS)
+
+#define RXMODE_BPF_VERDICT() (			\
+	RXMODE_BPF_VERDICT_INGRESS() ||		\
+	RXMODE_BPF_VERDICT_EGRESS())
+
+#define RXMODE_BPF() (				\
+	RXMODE_BPF_PASS() ||			\
+	RXMODE_BPF_VERDICT())
+
+#define RXMODE() (				\
+	RXMODE_NORMAL() ||			\
+	RXMODE_BPF())
+
+static struct socmap_ctx {
+	struct bench_sockmap_prog *skel;
+	enum SOCKMAP_ARG_FLAG mode;
+	#define c1	fds[0]
+	#define p1	fds[1]
+	#define c2	fds[2]
+	#define p2	fds[3]
+	#define sfd	fds[4]
+	int		fds[5];
+	long		send_calls;
+	long		read_calls;
+	long		prod_send;
+	long		user_read;
+	int		file_size;
+	int		delay_consumer;
+	int		prod_run_time;
+	int		strp_size;
+} ctx = {
+	.prod_send	= 0,
+	.user_read	= 0,
+	.file_size	= FILE_SIZE,
+	.mode		= ARG_FW_RX_VERDICT_EGRESS,
+	.fds		= {0},
+	.delay_consumer = 0,
+	.prod_run_time	= 0,
+	.strp_size	= 0,
+};
+
+static void bench_sockmap_prog_destroy(void)
+{
+	int i;
+
+	for (i = 0; i < sizeof(ctx.fds); i++) {
+		if (ctx.fds[0] > 0)
+			close(ctx.fds[i]);
+	}
+
+	bench_sockmap_prog__destroy(ctx.skel);
+}
+
+static void init_addr(struct sockaddr_storage *ss,
+		      socklen_t *len)
+{
+	struct sockaddr_in *addr4 = memset(ss, 0, sizeof(*ss));
+
+	addr4->sin_family = AF_INET;
+	addr4->sin_port = 0;
+	addr4->sin_addr.s_addr = htonl(INADDR_LOOPBACK);
+	*len = sizeof(*addr4);
+}
+
+static bool set_non_block(int fd, bool blocking)
+{
+	int flags = fcntl(fd, F_GETFL, 0);
+
+	if (flags == -1)
+		return false;
+	flags = blocking ? (flags | O_NONBLOCK) : (flags & ~O_NONBLOCK);
+	return (fcntl(fd, F_SETFL, flags) == 0);
+}
+
+static int create_pair(int *c, int *p, int type)
+{
+	struct sockaddr_storage addr;
+	int err, cfd, pfd;
+	socklen_t addr_len = sizeof(struct sockaddr_storage);
+
+	err = getsockname(ctx.sfd, (struct sockaddr *)&addr, &addr_len);
+	if (err) {
+		fprintf(stderr, "getsockname error %d\n", errno);
+		return err;
+	}
+	cfd = socket(AF_INET, type, 0);
+	if (cfd < 0) {
+		fprintf(stderr, "socket error %d\n", errno);
+		return err;
+	}
+
+	err = connect(cfd, (struct sockaddr *)&addr, addr_len);
+	if (err && errno != EINPROGRESS) {
+		fprintf(stderr, "connect error %d\n", errno);
+		return err;
+	}
+
+	pfd = accept(ctx.sfd, NULL, NULL);
+	if (pfd < 0) {
+		fprintf(stderr, "accept error %d\n", errno);
+		return err;
+	}
+	*c = cfd;
+	*p = pfd;
+	return 0;
+}
+
+static int create_sockets(void)
+{
+	struct sockaddr_storage addr;
+	int err, one = 1;
+	socklen_t addr_len;
+
+	init_addr(&addr, &addr_len);
+	ctx.sfd = socket(AF_INET, SOCK_STREAM, 0);
+	if (ctx.sfd < 0) {
+		fprintf(stderr, "socket error:%d\n", errno);
+		return ctx.sfd;
+	}
+	err = setsockopt(ctx.sfd, SOL_SOCKET, SO_REUSEPORT, &one, sizeof(one));
+	if (err) {
+		fprintf(stderr, "setsockopt error:%d\n", errno);
+		return err;
+	}
+
+	err = bind(ctx.sfd, (struct sockaddr *)&addr, addr_len);
+	if (err) {
+		fprintf(stderr, "bind error:%d\n", errno);
+		return err;
+	}
+
+	err = listen(ctx.sfd, SOMAXCONN);
+	if (err) {
+		fprintf(stderr, "listen error:%d\n", errno);
+		return err;
+	}
+
+	err = create_pair(&ctx.c1, &ctx.p1, SOCK_STREAM);
+	if (err) {
+		fprintf(stderr, "create_pair 1 error\n");
+		return err;
+	}
+
+	err = create_pair(&ctx.c2, &ctx.p2, SOCK_STREAM);
+	if (err) {
+		fprintf(stderr, "create_pair 2 error\n");
+		return err;
+	}
+	printf("create socket fd c1:%d p1:%d c2:%d p2:%d\n",
+	       ctx.c1, ctx.p1, ctx.c2, ctx.p2);
+	return 0;
+}
+
+static void validate(void)
+{
+	if (env.consumer_cnt != 2 || env.producer_cnt != 1 ||
+	    !env.affinity)
+		goto err;
+	return;
+err:
+	fprintf(stderr, "argument '-c 2 -p 1 -a' is necessary");
+	exit(1);
+}
+
+static int setup_rx_sockmap(void)
+{
+	int verdict, pass, parser, map;
+	int zero = 0, one = 1;
+	int err;
+
+	parser = bpf_program__fd(ctx.skel->progs.prog_skb_parser);
+	verdict = bpf_program__fd(ctx.skel->progs.prog_skb_verdict);
+	pass = bpf_program__fd(ctx.skel->progs.prog_skb_pass);
+	map = bpf_map__fd(ctx.skel->maps.sock_map_rx);
+
+	if (ctx.strp_size != 0) {
+		ctx.skel->bss->pkt_size = ctx.strp_size;
+		err = bpf_prog_attach(parser, map, BPF_SK_SKB_STREAM_PARSER, 0);
+		if (err)
+			return err;
+	}
+
+	if (RXMODE_BPF_VERDICT())
+		err = bpf_prog_attach(verdict, map, BPF_SK_SKB_STREAM_VERDICT, 0);
+	else if (RXMODE_BPF_PASS())
+		err = bpf_prog_attach(pass, map, BPF_SK_SKB_STREAM_VERDICT, 0);
+	if (err)
+		return err;
+
+	if (RXMODE_BPF_PASS())
+		return bpf_map_update_elem(map, &zero, &ctx.c2, BPF_NOEXIST);
+
+	err = bpf_map_update_elem(map, &zero, &ctx.p1, BPF_NOEXIST);
+	if (err < 0)
+		return err;
+
+	if (RXMODE_BPF_VERDICT_INGRESS()) {
+		ctx.skel->bss->verdict_dir = BPF_F_INGRESS;
+		err = bpf_map_update_elem(map, &one, &ctx.c2, BPF_NOEXIST);
+	} else {
+		err = bpf_map_update_elem(map, &one, &ctx.p2, BPF_NOEXIST);
+	}
+	if (err < 0)
+		return err;
+
+	return 0;
+}
+
+static int setup_tx_sockmap(void)
+{
+	int zero = 0, one = 1;
+	int prog, map;
+	int err;
+
+	map = bpf_map__fd(ctx.skel->maps.sock_map_tx);
+	prog = TXMODE_BPF_PASS() ?
+		bpf_program__fd(ctx.skel->progs.prog_skmsg_pass) :
+		bpf_program__fd(ctx.skel->progs.prog_skmsg_verdict);
+
+	err = bpf_prog_attach(prog, map, BPF_SK_MSG_VERDICT, 0);
+	if (err)
+		return err;
+
+	if (TXMODE_BPF_EGRESS()) {
+		err = bpf_map_update_elem(map, &zero, &ctx.p1, BPF_NOEXIST);
+		err |= bpf_map_update_elem(map, &one, &ctx.p2, BPF_NOEXIST);
+	} else {
+		ctx.skel->bss->verdict_dir = BPF_F_INGRESS;
+		err = bpf_map_update_elem(map, &zero, &ctx.p2, BPF_NOEXIST);
+		err |= bpf_map_update_elem(map, &one, &ctx.c2, BPF_NOEXIST);
+	}
+
+	if (err < 0)
+		return err;
+
+	return 0;
+}
+
+static void setup(void)
+{
+	int err;
+
+	ctx.skel = bench_sockmap_prog__open_and_load();
+	if (!ctx.skel) {
+		fprintf(stderr, "error loading skel\n");
+		exit(1);
+	}
+
+	if (create_sockets()) {
+		fprintf(stderr, "create_net_mode error\n");
+		goto err;
+	}
+
+	if (RXMODE_BPF()) {
+		err = setup_rx_sockmap();
+		if (err) {
+			fprintf(stderr, "setup_rx_sockmap error:%d\n", err);
+			goto err;
+		}
+	} else if (TXMODE_BPF()) {
+		err = setup_tx_sockmap();
+		if (err) {
+			fprintf(stderr, "setup_tx_sockmap error:%d\n", err);
+			goto err;
+		}
+	} else {
+		fprintf(stderr, "unknown sockmap bench mode: %d\n", ctx.mode);
+		goto err;
+	}
+
+	return;
+
+err:
+	bench_sockmap_prog_destroy();
+	exit(1);
+}
+
+static void measure(struct bench_res *res)
+{
+	res->drops = atomic_swap(&ctx.prod_send, 0);
+	res->hits = atomic_swap(&ctx.skel->bss->process_byte, 0);
+	res->false_hits = atomic_swap(&ctx.user_read, 0);
+	res->important_hits = atomic_swap(&ctx.send_calls, 0);
+	res->important_hits |= atomic_swap(&ctx.read_calls, 0) << 32;
+}
+
+static void verify_data(int *check_pos, char *buf, int rcv)
+{
+	for (int i = 0 ; i < rcv; i++) {
+		if (buf[i] != snd_data[(*check_pos) % DATA_REPEAT_SIZE]) {
+			fprintf(stderr, "verify data fail");
+			exit(1);
+		}
+		(*check_pos)++;
+		if (*check_pos >= FILE_SIZE)
+			*check_pos = 0;
+	}
+}
+
+static void *consumer(void *input)
+{
+	int rcv, sent;
+	int check_pos = 0;
+	int tid = (long)input;
+	int recv_buf_size = FILE_SIZE;
+	char *buf = malloc(recv_buf_size);
+	int delay_read = ctx.delay_consumer;
+
+	if (!buf) {
+		fprintf(stderr, "fail to init read buffer");
+		return NULL;
+	}
+
+	while (true) {
+		if (tid == 1) {
+			/* consumer 1 is unused for tx test and stream verdict test */
+			if (RXMODE_BPF() || TXMODE())
+				return NULL;
+			/* it's only for RX_NORMAL which service as reserve-proxy mode */
+			rcv = read(ctx.p1, buf, recv_buf_size);
+			if (rcv < 0) {
+				fprintf(stderr, "fail to read p1");
+				return NULL;
+			}
+
+			sent = send(ctx.p2, buf, recv_buf_size, 0);
+			if (sent < 0) {
+				fprintf(stderr, "fail to send p2");
+				return NULL;
+			}
+		} else {
+			if (delay_read != 0) {
+				if (delay_read < 0)
+					return NULL;
+				sleep(delay_read);
+				delay_read = 0;
+			}
+			/* read real endpoint by consumer 0 */
+			atomic_inc(&ctx.read_calls);
+			rcv = read(ctx.c2, buf, recv_buf_size);
+			if (rcv < 0 && errno != EAGAIN) {
+				fprintf(stderr, "%s fail to read c2 %d\n", __func__, errno);
+				return NULL;
+			}
+			verify_data(&check_pos, buf, rcv);
+			atomic_add(&ctx.user_read, rcv);
+		}
+	}
+
+	return NULL;
+}
+
+static void *producer(void *input)
+{
+	int off = 0, fp, need_sent, sent;
+	int file_size = ctx.file_size;
+	struct timespec ts1, ts2;
+	int target;
+	FILE *file;
+
+	file = tmpfile();
+	if (!file) {
+		fprintf(stderr, "create file for sendfile");
+		return NULL;
+	}
+
+	/* we need simple verify */
+	for (int i = 0; i < file_size; i++) {
+		if (fwrite(&snd_data[off], sizeof(char), 1, file) != 1) {
+			fprintf(stderr, "init tmpfile error");
+			return NULL;
+		}
+		if (++off >= sizeof(snd_data))
+			off = 0;
+	}
+	fflush(file);
+	fseek(file, 0, SEEK_SET);
+
+	fp = fileno(file);
+	need_sent = file_size;
+	clock_gettime(CLOCK_MONOTONIC, &ts1);
+
+	if (RXMODE_BPF_VERDICT())
+		target = ctx.c1;
+	else if (TXMODE_BPF_EGRESS())
+		target = ctx.p1;
+	else
+		target = ctx.p2;
+	set_non_block(target, true);
+	while (true) {
+		if (ctx.prod_run_time) {
+			clock_gettime(CLOCK_MONOTONIC, &ts2);
+			if (ts2.tv_sec - ts1.tv_sec > ctx.prod_run_time)
+				return NULL;
+		}
+
+		errno = 0;
+		atomic_inc(&ctx.send_calls);
+		sent = sendfile(target, fp, NULL, need_sent);
+		if (sent < 0) {
+			if (errno != EAGAIN && errno != ENOMEM && errno != ENOBUFS) {
+				fprintf(stderr, "sendfile return %d, errorno %d:%s\n",
+					sent, errno, strerror(errno));
+				return NULL;
+			}
+			continue;
+		} else if (sent < need_sent) {
+			need_sent -= sent;
+			atomic_add(&ctx.prod_send, sent);
+			continue;
+		}
+		atomic_add(&ctx.prod_send, need_sent);
+		need_sent = file_size;
+		lseek(fp, 0, SEEK_SET);
+	}
+
+	return NULL;
+}
+
+static void report_progress(int iter, struct bench_res *res, long delta_ns)
+{
+	double speed_mbs, prod_mbs, bpf_mbs, send_hz, read_hz;
+
+	prod_mbs = res->drops / 1000000.0 / (delta_ns / 1000000000.0);
+	speed_mbs = res->false_hits / 1000000.0 / (delta_ns / 1000000000.0);
+	bpf_mbs = res->hits / 1000000.0 / (delta_ns / 1000000000.0);
+	send_hz = (res->important_hits & 0xFFFFFFFF) / (delta_ns / 1000000000.0);
+	read_hz = (res->important_hits >> 32) / (delta_ns / 1000000000.0);
+
+	printf("Iter %3d (%7.3lfus): ",
+	       iter, (delta_ns - 1000000000) / 1000.0);
+	printf("Send Speed %8.3lf MB/s (%8.3lf calls/s), BPF Speed %8.3lf MB/s, "
+	       "Rcv Speed %8.3lf MB/s (%8.3lf calls/s)\n",
+	       prod_mbs, send_hz, bpf_mbs, speed_mbs, read_hz);
+}
+
+static void report_final(struct bench_res res[], int res_cnt)
+{
+	double verdict_mbs_mean = 0.0;
+	long verdict_total = 0;
+	int i;
+
+	for (i = 0; i < res_cnt; i++) {
+		verdict_mbs_mean += res[i].hits / 1000000.0 / (0.0 + res_cnt);
+		verdict_total += res[i].hits / 1000000.0;
+	}
+
+	printf("Summary: total trans %8.3lu MB \u00B1 %5.3lf MB/s\n",
+	       verdict_total, verdict_mbs_mean);
+}
+
+static const struct argp_option opts[] = {
+	{ "rx-normal", ARG_FW_RX_NORMAL, NULL, 0,
+		"simple reserve-proxy mode, no bfp enabled"},
+	{ "rx-pass", ARG_FW_RX_PASS, NULL, 0,
+		"run bpf prog but no redir applied"},
+	{ "rx-strp", ARG_CTL_RX_STRP, "Byte", 0,
+		"enable strparser and set the encapsulation size"},
+	{ "rx-verdict-egress", ARG_FW_RX_VERDICT_EGRESS, NULL, 0,
+		"forward data with bpf(stream verdict)"},
+	{ "rx-verdict-ingress", ARG_FW_RX_VERDICT_INGRESS, NULL, 0,
+		"forward data with bpf(stream verdict)"},
+	{ "tx-normal", ARG_FW_TX_NORMAL, NULL, 0,
+		"simple c-s mode, no bfp enabled"},
+	{ "tx-pass", ARG_FW_TX_PASS, NULL, 0,
+		"run bpf prog but no redir applied"},
+	{ "tx-verdict-ingress", ARG_FW_TX_VERDICT_INGRESS, NULL, 0,
+		"forward msg to ingress queue of another socket"},
+	{ "tx-verdict-egress", ARG_FW_TX_VERDICT_EGRESS, NULL, 0,
+		"forward msg to egress queue of another socket"},
+	{ "delay-consumer", ARG_CONSUMER_DELAY_TIME, "SEC", 0,
+		"delay consumer start"},
+	{ "producer-duration", ARG_PRODUCER_DURATION, "SEC", 0,
+		"producer duration"},
+	{},
+};
+
+static error_t parse_arg(int key, char *arg, struct argp_state *state)
+{
+	switch (key) {
+	case ARG_FW_RX_NORMAL...ARG_FW_TX_VERDICT_EGRESS:
+		ctx.mode = key;
+		break;
+	case ARG_CONSUMER_DELAY_TIME:
+		ctx.delay_consumer = strtol(arg, NULL, 10);
+		break;
+	case ARG_PRODUCER_DURATION:
+		ctx.prod_run_time = strtol(arg, NULL, 10);
+		break;
+	case ARG_CTL_RX_STRP:
+		ctx.strp_size = strtol(arg, NULL, 10);
+		break;
+	default:
+		return ARGP_ERR_UNKNOWN;
+	}
+
+	return 0;
+}
+
+/* exported into benchmark runner */
+const struct argp bench_sockmap_argp = {
+	.options	= opts,
+	.parser		= parse_arg,
+};
+
+/* Benchmark performance of creating bpf local storage  */
+const struct bench bench_sockmap = {
+	.name			= "sockmap",
+	.argp			= &bench_sockmap_argp,
+	.validate		= validate,
+	.setup			= setup,
+	.producer_thread	= producer,
+	.consumer_thread	= consumer,
+	.measure		= measure,
+	.report_progress	= report_progress,
+	.report_final		= report_final,
+};
--- a/tools/testing/selftests/bpf/progs/bench_sockmap_prog.c
+++ b/tools/testing/selftests/bpf/progs/bench_sockmap_prog.c
@@ -0,0 +1,65 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <linux/bpf.h>
+#include <bpf/bpf_helpers.h>
+#include <bpf/bpf_endian.h>
+
+long process_byte = 0;
+int  verdict_dir = 0;
+int  dropped = 0;
+int  pkt_size = 0;
+struct {
+	__uint(type, BPF_MAP_TYPE_SOCKMAP);
+	__uint(max_entries, 20);
+	__type(key, int);
+	__type(value, int);
+} sock_map_rx SEC(".maps");
+
+struct {
+	__uint(type, BPF_MAP_TYPE_SOCKMAP);
+	__uint(max_entries, 20);
+	__type(key, int);
+	__type(value, int);
+} sock_map_tx SEC(".maps");
+
+SEC("sk_skb/stream_parser")
+int prog_skb_parser(struct __sk_buff *skb)
+{
+	return pkt_size;
+}
+
+SEC("sk_skb/stream_verdict")
+int prog_skb_verdict(struct __sk_buff *skb)
+{
+	int one = 1;
+	int ret =  bpf_sk_redirect_map(skb, &sock_map_rx, one, verdict_dir);
+
+	if (ret == SK_DROP)
+		dropped++;
+	__sync_fetch_and_add(&process_byte, skb->len);
+	return ret;
+}
+
+SEC("sk_skb/stream_verdict")
+int prog_skb_pass(struct __sk_buff *skb)
+{
+	__sync_fetch_and_add(&process_byte, skb->len);
+	return SK_PASS;
+}
+
+SEC("sk_msg")
+int prog_skmsg_verdict(struct sk_msg_md *msg)
+{
+	int one = 1;
+
+	__sync_fetch_and_add(&process_byte, msg->size);
+	return bpf_msg_redirect_map(msg, &sock_map_tx, one, verdict_dir);
+}
+
+SEC("sk_msg")
+int prog_skmsg_pass(struct sk_msg_md *msg)
+{
+	__sync_fetch_and_add(&process_byte, msg->size);
+	return SK_PASS;
+}
+
+char _license[] SEC("license") = "GPL";